Two fast algorithms for computation of h-index

Abstract

Despite its many limitations, h-index is one of the popular productivity indicators used by many scholarly indexing databases and search engine tools. Apart from indicating scientific and inventive productivity at individual level, it can be used for other actors such as journals, institutions, countries and assignees. Though h-index is relatively simple to compute than most of the h-type indicators for a single actor profile, efficient algorithms are always useful to manage large databases or repositories that require handling and frequent updating of large number of such profiles. However, there are very few attempts to improve the computation of h-index. In this work, we introduce two new estimation-based algorithms for computation of h-index on sorted profile and compare it with two existing approaches using (1) the real dataset of scholarly profiles of 50 eminent researchers in the area of information science and scientometrics, (2) a real dataset of 4177 scholars with h-index greater than 100 and (3) three sets of simulated profiles of actors. Both the algorithms are found to be performing better than their existing counterparts. Furthermore, we attempt to provide guidelines for the choice of strategies to implement these algorithms to ensure maximisation of speed.